github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-23 14:13:13 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #12141] Models are not running on GPU, only using CPU #33830

New Issue
Closed
opened 2026-04-22 16:54:56 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-22 16:54:56 -05:00
Owner
Copy Link

Originally created by @iloveumyfriend on GitHub (Sep 1, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12141

What is the issue?

Hello,

I am running Ollama with a GPU available (NVIDIA, confirmed via nvidia-smi), but models seem to only use the CPU.

  • htop shows 100% CPU usage
  • nvidia-smi shows that Ollama is started, but no significant GPU utilization
  • Logs show lines like:
    offloading output layer to CPU
    offloaded 0/49 layers to GPU

Environment:

  • Ollama version: ollama version is 0.11.6
  • GPU: NVIDIA H100 80GB
  • OS: Ubuntu - 24.04.2

nvidia-smi.txt

Relevant log output

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.11.6

Originally created by @iloveumyfriend on GitHub (Sep 1, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12141 ### What is the issue? Hello, I am running Ollama with a GPU available (NVIDIA, confirmed via nvidia-smi), but models seem to only use the CPU. - htop shows 100% CPU usage - nvidia-smi shows that Ollama is started, but no significant GPU utilization - Logs show lines like: `offloading output layer to CPU` `offloaded 0/49 layers to GPU` Environment: - Ollama version: `ollama version is 0.11.6` - GPU: NVIDIA H100 80GB - OS: Ubuntu - 24.04.2 [nvidia-smi.txt](https://github.com/user-attachments/files/22074736/nvidia-smi.txt) ### Relevant log output ```shell ``` ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.11.6
GiteaMirror added the bug label 2026-04-22 16:54:56 -05:00
GiteaMirror commented 2026-04-22 16:54:59 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 1, 2025):

Add the full server log.

<!-- gh-comment-id:3241484505 --> @rick-github commented on GitHub (Sep 1, 2025): Add the full server log.
GiteaMirror commented 2026-04-22 16:54:59 -05:00
Author
Owner
Copy Link

@iloveumyfriend commented on GitHub (Sep 1, 2025):

journalctl.txt

<!-- gh-comment-id:3241500855 --> @iloveumyfriend commented on GitHub (Sep 1, 2025): [journalctl.txt](https://github.com/user-attachments/files/22074865/journalctl.txt)
GiteaMirror commented 2026-04-22 16:55:00 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 1, 2025):

Sep 01 10:15:07 AI ollama[2128]: time=2025-09-01T10:15:07.233+03:00 level=INFO source=server.go:531 msg=offload library=cpu

Despite detecting the GPU, the server thinks that the only usable backend is the CPU. The start of the log would show device detection but unfortunately has not been included. Do the following and post the output:

sudo systemctl stop ollama
sudo systemctl start ollama
ollama run gpt-oss:20b ''
journalctl -u ollama --no-pager --since "$(systemctl show ollama --property=ActiveEnterTimestamp --value)"
<!-- gh-comment-id:3241710537 --> @rick-github commented on GitHub (Sep 1, 2025): ``` Sep 01 10:15:07 AI ollama[2128]: time=2025-09-01T10:15:07.233+03:00 level=INFO source=server.go:531 msg=offload library=cpu ``` Despite detecting the GPU, the server thinks that the only usable backend is the CPU. The start of the log would show device detection but unfortunately has not been included. Do the following and post the output: ``` sudo systemctl stop ollama sudo systemctl start ollama ollama run gpt-oss:20b '' journalctl -u ollama --no-pager --since "$(systemctl show ollama --property=ActiveEnterTimestamp --value)" ```
GiteaMirror commented 2026-04-22 16:55:00 -05:00
Author
Owner
Copy Link

@iloveumyfriend commented on GitHub (Sep 1, 2025):

Thank you for the instructions. After stopping and restarting the Ollama service, the models started running on the GPU as expected. It seems the issue was resolved after restarting.

Do you know what could have caused Ollama to initially fall back to CPU even though the GPU was detected?

journalctl.txt

<!-- gh-comment-id:3241775279 --> @iloveumyfriend commented on GitHub (Sep 1, 2025): Thank you for the instructions. After stopping and restarting the Ollama service, the models started running on the GPU as expected. It seems the issue was resolved after restarting. Do you know what could have caused Ollama to initially fall back to CPU even though the GPU was detected? [journalctl.txt](https://github.com/user-attachments/files/22076293/journalctl.txt)
GiteaMirror commented 2026-04-22 16:55:01 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 1, 2025):

Do you know what could have caused Ollama to initially fall back to CPU even though the GPU was detected?

No idea. As mentioned, the previous log may include relevant details.

<!-- gh-comment-id:3241793158 --> @rick-github commented on GitHub (Sep 1, 2025): > Do you know what could have caused Ollama to initially fall back to CPU even though the GPU was detected? No idea. As mentioned, the previous log may include relevant details.
GiteaMirror commented 2026-04-22 16:55:02 -05:00
Author
Owner
Copy Link

@iloveumyfriend commented on GitHub (Sep 1, 2025):

Thank you anyway for your help! I’ll keep an eye on the logs in case this happens again.

<!-- gh-comment-id:3241807847 --> @iloveumyfriend commented on GitHub (Sep 1, 2025): Thank you anyway for your help! I’ll keep an eye on the logs in case this happens again.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#33830
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?